Goose v1.48.0 Changelog

This release is all about giving you more options and making the app safer and easier to use. It matters to anyone who uses Goose to build code or automate tasks, especially if you like switching between different AI models or need a reliable tool that won't leak sensitive info.

What's new

• More AI Providers: You can now easily connect to new AI services like TrustedRouter, OpenCode Zen, Gondola, SayGM, and Lynkr. You also get access to updated models like Z.ai GLM-5.3 and OpenAI GPT-5.6.
• Built-in Web Skills: Goose now has built-in "skills" for web searching and browser use, meaning the AI can go online to find answers or interact with websites for you.
• Smarter Hooks: If you use automated "hooks" (custom scripts that run before the AI uses a tool), you can now set them to do something specific if they fail.
• Desktop Upgrades: When scheduling tasks, you can pick from saved recipes. The chat history is now neatly grouped into collapsible menus, and there is a handy Git branch indicator at the bottom of your chat.
• Fresh Command Line Tools: If you use the text-based command line, you can now type /new to instantly start a fresh chat without restarting the whole program. You can also give nicknames to your extensions when you load them.

Improvements

• Better Cost Tracking: If you use custom AI providers, you can now set up specific fields to track exactly how much money you are spending.
• Easier Provider Sorting: Providers you have already set up will now automatically show up at the very top of the list, so you don't have to hunt for them.
• Smarter Chat Input: The chat box now automatically focuses itself when you start typing, and the model picker remembers which AI models you used most recently.
• Better System Monitoring: For developers tracking how the app behaves, Goose now sends more detailed telemetry data (info about how the app is running) and forwards session IDs to help trace issues.

Fixes

• Major Security Tightening: A huge number of security holes were fixed. For example, the app now safely blocks things if it sees weird, invisible Unicode text (special characters that can sometimes trick systems). It also makes sure to hide sensitive passwords and website links if an error happens, and it stops endless loops that could crash the system.
• Provider Settings Saving: Fixed a bug where saving your provider settings would accidentally overwrite your custom setup with default settings.
• Smarter Error Messages: When you hit a limit on how much text you can send to an AI, the app now correctly tells you "context length exceeded" instead of giving a confusing error.
• Cheaper Fast Calls: Fixed an issue where quick, one-off questions to fast AI models were accidentally paying extra fees for prompt caching.
• Cleaner AI "Thinking": Fixed how the app handles AI models that "think" before answering, making sure those thought blocks don't get messy or break the stream of text.

Breaking changes

No breaking changes in this release.